What problem does it solve? Agents lose everything they learned when a session ends, forcing every future session to re-derive project conventions, constraints, and architectural decisions from scratch. This Skill teaches an agent to use the Canopy Project Intelligence Layer (PIL) so knowledge outlives individual sessions. ## Core Features & Use Cases - Session-start context pull: Call get_tools(scope="session_start") to receive the workspace brief, and intelligence_get_context(scope="full") for deep architecture work or onboarding. - Durable knowledge persistence: Register facts (conventions, constraints, configuration truths) and patterns (recurring code structures, workflow rules, architectural decisions) with intelligence_upsert so future sessions inherit them. - Session closure protocol: Call get_tools(scope="close_session") to upsert a session summary and report workspace status before ending. - Use Case: While exploring a Rust codebase, you discover that all DB tables use snake_case with an _at suffix. You upsert it as a fact immediately, and the next agent session starts already knowing the convention. ## Quick Start Ask the agent to pull the Canopy session-start context and upsert any durable conventions it discovers during this session.
